compared the proportion (percentage) of families of different sizes in their two test groups and in the control and by this method found no evidence of significantly increased risk in children from larger families either of malnutrition or of death from diseases associated with malnutrition.
They concluded that "The size of the family appears to have little effect per se on the health of the younger members, families being 'at risk' because of poverty, accompanied by poor parental capacity" (16, p. 893).
Because, as noted, there are few such studies and the findings of Robertson and Kemp in this study are so different from those of others, a methodological problem involved here deserves careful scrutiny. In treating their data they compared the distribution of differing numbers of children per family in three groups of families, each selected quite differently: the first, because there was a malnourished preschool child in the family, the second because a child in the family died of a specified cause, and the third because the mother delivered a newborn infant during a defined period of time. What is needed, however, if we are to understand the effect of family size is a comparison of the prevalence of malnutrition or of death rates in a substantial population of children grouped according to family size. Only then can a valid statement be made concerning the presence or absence of increased risk associated with family size.
This point has been spelled out by Chen and Cobb in their paper, "Family Structure in Relation to Health and Disease" (17), in which they state:
With regard to sampling method, the most important thing is to check that the sampling procedure is the same for cases and controls. It is, for example, wholly inappropriate to compare the distribution of family sizes among patients with the distribution reported in a census. The reason for this is evident on consideration of an hypothetical population consisting of equal numbers of families with one child and ten children. If one picks families and asks how many children there are, one has an equal chance of picking large families or small families and concludes that the average size is 5.5 children. If, on the other hand, one picks individuals and asks them how many siblings they have, the probability of picking a person from a large family is ten times as great as the probability of picking one from a small family. In this situation, one would conclude that average size of family is 9.2 children. . . .
The soundest way to approach this entire problem is, of course, to examine attack rates of disease by size of sibship, but this can only be undertaken when a survey of a total population or sample of the population has been conducted. (17, p. 549)
Family Size and Mortality. Morbidity, or illness, is but a continuum which, in the end, may result in death. There is an abundance of data concerning the association between family size and mortality rates. In one study of a rural